Individual and age information in the gigglenote. A. Position of individuals' centroids in function of the first three discriminant variables that maximizes individual separation. The left panel shows the centroids on a 2D projection for discriminant variables 1 and 2 and the right panel shows the centroids on a 2D projection for discriminant variables 1 and 3. The discriminant functions have been scaled so that the within-variance is 1; in other words, assuming normality, for each animal, 67% of its notes would be found in a sphere of radius 1 centred on these centroids. The plotted ovals around the centroids show one standard error of the mean. B. The confusion matrix obtained from the DFA on the cross-validation data set. The confusion matrix shows by colouring cell (i, j) the conditional probability of guessing that the test giggle notes came from individual j when in fact it was emitted by i. C. Average (cumulative) percent correct of calls classification according to the emitter's identity as a function of the number of random iterations for the data used to fit the discriminant functions (left) and for the data used for cross-validation (right). The dotted lines show two standard errors deviations from the final mean. The red line shows chance. D. Correlation between age and giggles' acoustic structure as described by the first discriminant function (left panel) and as described by fundamental frequency (right panel).
Mathevon et al. BMC Ecology 2010 10:9 doi:10.1186/1472-6785-10-9